As far as I have been able to determine the owner of the website is a member here. https://www.diyaudio.com/community/members/sampler.124579/
Interesting, for LED and zener, corresponds pretty much to my observations in my bench (the silent and noisy ones).
Having both Christers and MV’s practical measurements is really great, but whenever you have two sets of data the natural temptation is to compare them…
Christer’s original measurements for a 12V zener (BZX55/C12) operating at 5mA shows what understand to be a total noise level of about 0.3uV rms measured within his sound card’s bandwidth. Assuming this is around 20KHz and the zener noise spectrum is flat, then Christer’s measurements would imply a spectral noise density of roughly 0.3uV/sqrt(20K) or ~2nV/rtHz.
However, MV’s measurements for a 13V zener also operating at 5mA appear show a flat spectral noise density of roughly 330nV/rtHz which is a lot higher.
This seems an odd result as I’d have thought that 12V and 13V zeners would have had broadly similar noise behaviour.
I wonder if anyone might clarify this for me?
Some things to take into consideration when comparing the articles
1) FFTs -- the bin and windowing methods should be stated.
2) In Scott's article C2 = 1uF. If you port this value over to the schematic as shown, the response below 10Hz should be discarded. Not that important for audio, but for measuring 1/f noise it matters (oscillators, voltage references etc). Use 10uF
3)Nicads are quieter vis-a-vis alkalines
4) Scott used a rail-to-rail opamp, AD823. This neatly removes one noise source, V-
1) FFTs -- the bin and windowing methods should be stated.
2) In Scott's article C2 = 1uF. If you port this value over to the schematic as shown, the response below 10Hz should be discarded. Not that important for audio, but for measuring 1/f noise it matters (oscillators, voltage references etc). Use 10uF
3)Nicads are quieter vis-a-vis alkalines
4) Scott used a rail-to-rail opamp, AD823. This neatly removes one noise source, V-
Thanks very much for these sensible points regarding the noise measurements.
However, given that both articles demonstrate that the measurement noise floor is below the Zener noise, I’m not sure they explain the significant difference (~2 orders of magnitude) in the measured noise for the 12V and 13V Zeners?
Looking at the original noise measurements 12V Zener has about 1 decade lower noise than e.g. 5.6V Zener (0.3uV vs. 2.9uV). IMO this is an indication of a measurement error so I would disregard the original zener noise measurements.
There is a transition from the Zener effect to an avalanche breakdown mechanism that occurs at around 5V to 8V depending upon the details of the device structure as explained here: https://www.onsemi.com/pub/collateral/hbd854-d.pdf.
As a result one might expect the noise behaviour of Zeners with breakdown voltages of around 6V, where both mechanisms are operating simultaneously, to be rather different to Zeners with lower or higher breakdown voltages where only one mechanism is dominant.
Indeed, this was something that Christer explicitly intended to investigate in his original measurements and why he included Zeners of 2.7V, 5.6V, 6.8V and 12V. His results showed that the 5.6V and 6.8V diodes had higher noise levels than either the 2.7V and 12V devices.
So I would be reluctant to disregard Christer’s noise measurements due to the variation in 5.6V and 12V Zener noise levels.
Having looked at various manufacturers’ data, I’m leaning to the view that because Zener noise appears to be neither specified nor measured by manufacturers, it’s not a control parameter and hence one should probably just expect a wide variation in noise from different devices and manufacturers.
As a result one might expect the noise behaviour of Zeners with breakdown voltages of around 6V, where both mechanisms are operating simultaneously, to be rather different to Zeners with lower or higher breakdown voltages where only one mechanism is dominant.
Indeed, this was something that Christer explicitly intended to investigate in his original measurements and why he included Zeners of 2.7V, 5.6V, 6.8V and 12V. His results showed that the 5.6V and 6.8V diodes had higher noise levels than either the 2.7V and 12V devices.
So I would be reluctant to disregard Christer’s noise measurements due to the variation in 5.6V and 12V Zener noise levels.
Having looked at various manufacturers’ data, I’m leaning to the view that because Zener noise appears to be neither specified nor measured by manufacturers, it’s not a control parameter and hence one should probably just expect a wide variation in noise from different devices and manufacturers.
Yes, I'm aware of the different breakdown mechanisms. In the original measurements 12V zener has much lower noise than even the 2.7V zener which is well into the predominantly zener breakdown region.
According to e.g. this paper 12V zeners are at least 15-20dB noisier than e.g. 5V zeners.
https://www.analog.com/en/resources/design-notes/building-a-lowcost-whitenoise-generator.html
According to e.g. this paper 12V zeners are at least 15-20dB noisier than e.g. 5V zeners.
https://www.analog.com/en/resources/design-notes/building-a-lowcost-whitenoise-generator.html
@bohrok2610 Many thanks for the reply and link. There are some interesting observations in the paper:
However, MV’s measurements appear to show a significant variation in noise level with current?
Which is bit of a nuisance!